Agent-based modeling is a type of modeling in which
the action and interactions of autonomous agents.
Both with each other and the environment
are explicitly modeled in computer program.
The agents can be nearly anything,
animals, people, cities, nations.
Usually the agents follow simple rules
and are influenced by the other agents
and their immediate surroundings.
Let’s use an example to show what this means.
In this model there is some fish,
the agents that follow one simple rule,
swim straight.
When a fish exits on one side it,
re-enters on the opposite side,
not very excited right.
To make things more interesting,
let’s add a second rule.
Occasionally each fish randomly changes its direction
a little bit away from its current direction.
This is more interesting, but still not very exciting.
The fish aren’t interacting with each other.
Adding one final, third rule makes all the difference.
Rather than going more or less in the same direction,
each fish moves in more or less the average direction
of all the fish in its local neighborhood.
This mimics real life, an actual fish
in an actual school can’t see the entire school,
but it can sense the fish closed by.
Agent-based modeling is much more flexible
than other types of modeling
such as equation-based modeling.
With enough rules, almost any behavior can be modeled
in any phenomenon can be observed.
But this flexibility also has its down side.
Agent-based models with too many rules are hard to understand.
If there are twenty rules effecting one agent
how are we to know which rule is felt most strongly.
For example what rules are necessary to observe
a phenomenon such as the schooling of fish
could a simpler model yield schooling.
Nevertheless a well-designed agent-based model
can be very informative.
It can tell us which combinations of behavioral rules
for agents and environmental conditions yield
interesting behavior such as the schooling of fish.
What’s cool about this model
is that the fish actually forms school
based on three simple rules.
They present a different way to understand
how phenomena and patterns can arise
from very simple behavior,
which is a hallmark of complex systems.
Agent-based modeling is foundational to the science
that we do here at the Santa Fe Institute.